Abstract:
Background
Although much progress has been made in early detection, diagnosis and therapy of breast cancer in the past few years, it is still one of the leading causes of cancer death among women. Mammography is the mainstay of early diagnosis of breast cancer. Microcalcifications in mammograms play a very important role in the detecting of breast cancer, especially early stage; however they are not breast cancer specific. Studies report, that mammography show microcalcification in up to nearly 50% of cases of breast cancer. The BI-RADS Lexicon defines parameters to classify the mammographic features into standardized diagnostic categories. Yet, the classification of microcalcification raises difficulties, particularly because there are no recommendations in the Lexicon on how to link the descriptive findings to the BI-RADS assessment categories.
Purpose
The purpose of our study is to evaluate a new kind of classification scheme of microcalcification. This newly developed matrix-like scheme combines the morphology and distribution parameters of microcalcification and gives for each combination an assignment for the BI-RADS assessment categories.
Material and Methods
The digital mammography images of 203 patients with microcalcification were twice interpreted by 4 Readers at intervals of at least four weeks. They classified the images on the one hand according the classification scheme and on the other hand according to the BI-RADS categories.
The classification process with the ‘matrix’ scheme divides the morphology into three groups: typically benign, intermediate and typically malignant calcifications. The distribution is divided into five groups: diffuse, round or oval, regional, segmental and linear.
All patients underwent vacuum core biopsy or open surgery. The histological findings were compared to the two radiological classifications. BI-RADS categories 2 and 3 were considered as representing benign cases, whereas BI-RADS categories 4 and 5 were counted as malignant cases. The significance of the correlation between the Bi-RADS classification and the classification with the matrix scheme and the diagnostic indices (sensitivity, specificity, positive predictive value) were calculated using standard statistical methods.
Results
The most frequent histological results were observed in 124 malignant cases in form of DCIS and invasive ductal cancer. The most frequent benign diagnoses were fibrocystic changes and adenosis.
The classification according to BI-RADS, as well as the classification according to the matrix scheme, showed high inter- and intraobserver variability.
The sensitivity and specificity showed significant differences. For the classification with the matrix-scheme the results are 81% and 54%. The results of the classification according to BI-RADS are 86 % and 52 %. In the case of BI-RADS categories 2/3/4/5, malignant diagnoses were found in 13 %/24 %/ 53 %/86 % cases with the classification according BI-RADS. The classification with the matrix scheme found 8 %/32 %/48 %/84 % were malignant cases.
The matrix scheme shows an increase of malignant cases from diffuse to linear distribution and from typically benign to typically malignant morphology.
Discussion
The results of our study show that the complexity of microcalcification and the difficulties in detecting and describing them still remain the main problem when interpreting mammograms. The classification with the matrix scheme does not significantly reduce the observer variability neither does it improve the results of sensitivity and specificity. It is still a difficult question of how to link the descriptive findings to the assessment categories. Further evaluation regarding this task and the assignment of the matrix scheme will be necessary.